A power amplifier linearization algorithm for indirect learning architecture

To address the shortcomings of the current linearization algorithm based on indirect learning architecture (ILA) that power amplifier compensation of non-linear is poor, as well as its spectral distortion improvement is not obvious, an alternating iteration algorithm based on power detection module is proposed. The algorithm adopts the data window to intercept the ILA of the post-distortion feedback back to the power of the signal stream and through the power detection module to calculate its power, filter the power of the largest signal data stream, according to the number of iterations, so that its output is switched between high-power signal stream and random signal stream and sent to the post-distortion module for training. The pre-distortion module and post-distortion module both adopt the memory polynomial model. LTE signal with a peak-to-average ratio of 9 dB is used to simulate the real GaN amplifier through the online test platform RF WebLab. The simulation results show that the out-of-band rejection effect of the proposed algorithm is optimized by about 5 dB and 2.5 dB compared with the conventional sequential data stream processing algorithm and high-power data stream processing algorithm, and the average normalized mean square error is optimized by about 1 dB and 0.6 dB, respectively.